Abstract | ||
---|---|---|
Recently, 3D hand pose estimation methods taking point cloud as input show the most advanced performance. We present a new 3D deep learning hand pose estimation network for an unordered point cloud. Our approach utilizes EdgeConv layer as the basic element, where an attention embedding version EdgeConv layer is proposed for feature extraction in hand pose estimation task. To improve the result, we design a hand pose improvement network that inputs points whose are in the neighbor of the estimated fingers and outputs a rectify hand pose. We evaluate our method on several famous datasets to prove that our method can get excellent result compared to some most advanced methods. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1007/978-3-030-22796-8_10 | ADVANCES IN NEURAL NETWORKS - ISNN 2019, PT I |
Keywords | Field | DocType |
3D hand pose estimation, Point cloud, Attention embedding module | Graph,Embedding,Pattern recognition,Computer science,Feature extraction,Pose,Artificial intelligence,Deep learning,Point cloud | Conference |
Volume | ISSN | Citations |
11554 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 0 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Xu Jiang | 1 | 6 | 3.46 |
Xiaohong Ma | 2 | 21 | 9.12 |